Usage of the graph clustering algorithm to the recognition of geotechnical layers

The aim of the paper is to present the approach to the application of the graph clustering algorithm to the recognition of geotechnical layers from the dilatometer tests. Results of the measurements obtained from the DMT test in the test site (subsoil of one of the buildings in the Warsaw University of Life Sciences campus) were analyzed by the clustering algorithm which was able to extract the separate groups of the measurements, representing identical soil type. This method is parameterized, so its verification by the geotechnical experts was necessary to determine the optimal parameter values. They lead to the determination of the soil types as close to the actual situation, as possible. Also, the output of the algorithm was analyzed by the geotechnical experts to identify and label the extracted soil types.